Some decisive features have been added to the OMax computer assisted improvisation system: one is a video interface, which makes the virtual clone of the performer not only audible but visible. Another one is a drastic extension of the symbolic world on which the improvisation is built: to an event-driven (basically pianoroll-like) representation has been added another one using spectral descriptors. The results are particularly exciting for a number of musical cases (timbre oriented music of course, but not only). It adds to the generality of an already agnostic system. The relationship of OMax to real-time, especially in follow mode, has to be noted in case of a music performed on a network. As for the video features, they allow a kind of sound driven total performance, in which written parts (with eventual filmed pictures) can be the pretext to free improvisatory moments intimately related to them. The fact OMax can continue an existing session (improvised or written) opens the door to all kind of hybrid interconnections between what is written, what is improvised, what is already existing and what is to be.